PROYECTO RESTAURACIÓN 1990

A QUANTITATIVE ASSESSMENT OF RUST SUSCEPTIBILITY IN VARIETIES OF ANTIRRHINUM MAJUS

INTRODUCTION

In many antirrhinum trials (Peltier, 1919; Doran, 1921;

Beaumont and Stanland, 1935; Fikry, 1939) the susceptibility of varieties to rust was only assessed once, generally at the end of the season and whilst a number of varietal differences were reported the comparisons were mainly qualitative. Since these early trials there have been many advances in the science of epidemiology. Plant pathologists have recognised four important components of an epidemic (Horsfall and Cowling, 1977);

i a susceptible host

ii a virulent and aggressive pathogen iii a favourable environment

iV time

Factors i to iii must be near their optimum for a period of time for the epidemic to flourish. Barratt (1945) was the first to recognise the importance of the rate of disease increase and plotted "disease trend curves" for the field performance of fungicides on tomato

varieties. Large (1952) used successive disease assessments to depict the progress of blight on various crops. He found the characteristic disease progress curve was sigmoid and this has been confirmed by many other workers. The disease development is initially slow because there is a relatively small amount of inoculum, it then accelerates and slows down again when there is little healthy host tissue left for the

pathogen to colonize.

Van der Plank (1963) has recognised three phases of an epidemic based on the sigmoid curve.

(1) The logarithmic phase where the increase of the pathogen is unhindered by the overlapping of lesions and is independent of

the amount of disease present. This phase has also been called the exponential phase by Zadoks and Schein (1979) and extends until theanount of disease (X) ■ 0*05.

(2) The logistic phase extends until X ■ 0*35 (Van der Plank, 1963) and until X ■ 0*50 (Zadoks and Schein, 1979).

(3) The terminal phase begins when X is greater than 0*50. It is during this phase that most damage iscfane to the crop.

It is generally agreed that the disease progress curves need to be transformed for subsequent analysis and in recent years there have been many attempts to fit data of this kind to various mathematical equations.

There does not appear to be a single transformation that is satisfactory for all plant diseases, thus semi— log, log-log, probit and logit

transformations have all been proposed. The results of all these

transformations must be interpreted with care since the rate of disease progress may vary considerably during an epidemic and this may be

disguised by a straight line transformation (Kranz, 1977).

Van der Plank (1963) proposed a logistic transformation

(loQg ( X/ (l-X) ) against time) for "compound interest" disease. This type of disease increase has been called polycyclic by Zadoks and Schein (1979) because the pathogen multiplies through successive generations during the course of an epidemic. The correction factor (l-X) is included to allow for the decreasing proportion of tissue left for infection. The logistic transformation fits a number of fungal

epidemics quite well (Simmonds, 1979) and the use of this transformation of the disease progress curve for an epidemic of Puccinia antirrhini on varieties of antirrhinum is discussed in Chapter 6.

An estimate of r, the apparent infection rate (Van der Plank, 1963, 1968, 1975) may be derived from the logistic transformation. The use of r has been critised because it varies with the stage of development of the plant and progress of epidemic (Parlevliet, 1979), the environment (Waggoner, 1965), and it does not provide a test of significance

(Kranz, 1974$ but despite these disadvantages r has been used for a number of purposes in epidemiological research:

to compare epidemics in a range of cultivars (Fry, 1978);

to assess the effects of fungicide on disease progress (Kannwischer and Mitchell, 1978);

to assess loss of yield (James, Callbeck, Hodgson and Shih, 1971);

to compare sanitation methods (Berger, 1977);

to assess the effect of plant spacing (Strandberg and White, 1978);

to predict the future course of an epidemic (Merrill, 1967).

An epidemic of rust on varieties of antirrhinum was studied to establish whether the susceptibility of previously "rust-resistant"

varieties was due to the loss of genetic resistance in Antirrhinum majus.

In addition it was thought that some varieties exhibited more "field resistance" to the rust. If these varieties could be identified they

MATERIALS AND METHODS

1. Cultivation

During the summers of 1978 and 1979 one hundred and thirty-one varieties of A. majus were tested for their susceptibility to the rust fungus at two locations in Surrey, England; Royal Holloway College, Egham and The Royal Horticultural Society's Garden at Wisley. The varieties were listed with the source of seed in Appendix 4.1 and the addresses of contributors given in Appendix 4.2. The very

susceptible variety, Malmaison (No.67) was used to ensure that there was a uniform high infection over the plots and as a control to compare the degree of rust infection at the two locations.

The varieties number 1 - 67 were included in trials of 1978. The seed was sown on 27th February, seedlings pricked out between the

15th and 30th March and the young plants planted out on the 17th and 18th May at Royal Holloway College and the 30th May at Wisley.

Both the plots were cleared in the late season of 1978 and left for the winter months. In April a general fertilizer "Growmore" was applied at the rate of 4 oz per square yard (l60g per square metre) and the plots rotivated to a depth of 4 ins. (lOcm).

The trials held during the summer of 1979 included the control variety No. 57, ten varieties repeated from the trial in 1978 - Nos. 1, 15, 18, 29, 37, 39, 45, 52, 56, 62 to assess the seasonal differences and varieties numbered 68 to 131 inclusive. The seed was sown on the 13th March, pricked out between 27th March and 6th April and planted out on 15th and 16th May at Royal Holloway College and on the 4th June at Wisley.

2. Design of Experiment.

At each location a plot measuring 25 metres x 15 metres was used for the trials. The two plots were treated as replicates and were therefore planted in exactly the same arrangement. Basically the varieties were planted in units of three rows (Fig.4.1). A complete row of "Malmaison" was planted in the middle to ensure all areas of the plot were exposed to high inoculum of rust spores. A row of test plants were planted either side of each row of "Malmaison". Each unit of three rows was separated by a path. In addition two guard rows of

"Malmaison were planted round the perimeter of the plot.

Figure 4,1 Arrangement of rows of the control and test varieties

• in the' Antirrhinum Trials

PATH

row of test varieties

— row of control variety - Malmaison row of test varieties

Unit of three rows

PATH

In the trial held during the summer of 1978 the varieties were divided into three height categories; dwarf - Nos. 1 - 13, intermediate Nos. 25 — 56 and tall — Nos. 14 - 24. The spacing between the rows were increased from 0*3 metres in the dwarf section, to 0*4 metres in

the intermediate section to 0*5 metres in the tall section, whereas the distance between plants within each row was constant at 20cm.

Each variety was planted in three blocks (a, b and c) of nine plants in a completely randomized design within each section. In addition, fourteen blocks of the control variety, "Malmaison", were planted within the rows of test plants in a stratified random design in order

to determine whether there was uniform rust infection throughout the plot. The resulting arrangement of varieties included in the plot for 1978 is shown in Figure 4.2. The complete rows and guard rows of

"Malmaison" are shown by the thicker lines.

Since there was no apparent correlation between the height of the plants and the amount of ground covered it was decided to keep the spacing between the rows constant in 1979. The rows were planted D*4m apart and the paths were 0*7m wide. This arrangement permitted the use of a randomized block design which is consistently more

accurate because the restrictions in the design reduce the experimental errors. (Cochran and Cox, 1958; Fisher, 1948). The plot was divided into three sections (a, b and c) and each variety placed once at random within each section. Thus, each variety was represented by three blocks of nine plants as in 1978. Nine blocks of "Malmaison", three in each section, were planted for the reason given above. In addition two guard rows of "Malmaison" surrounded the plot. The arrangement of varieties in the 1979 plot is shown in Figure 4.3.

3. Disease Assessment

Disease is generally assessed either as;

i disease incidence - expressed by the number of plants infected as a percentage of the total or

Figure 4,2 1978 Trial - Arrangement of varieties on plot

l a 1 9a 3a l b 5a 1 2 a 6a 6a 5b 11a 67a 3b

2a

5b 1 3 a 9b 4a 8b 10a ?a 2b 57b 4b 9c 10b

67 c 7b 8c 10c 12b 1 3 b 3c 11b 12c 7c 11c 1 3 c 2 c

4c 6 c 5c

39a 31a 32a

52a 52a 67d 1 c

55a

^{45 a}

51a

46a 6 éa 40a 35a 45b 43a 35a

25a

6 2 b 39a 25a

65a

34a_ 5 0 a

32b

57e 44 a

57a

^{50a 29a 56b} 31b

37a 4 8 a 28 a 4 1 a 45b 2 8 b 48 b 6l a 41b 38a 59a 55a

51b 55b 87 f 50b 27a 43b 876

63a 35b

25b 47 a 60b

4 9 a 55b 55c

58 b

44b

33a 25b

42a

53a

28 c 44c 5 8 c

65b 49b 47b 54a 37b 51b 47c 42b 5 2 b 30b 67h

55c

⁵⁷¹

4 1 c 5 1 c 43c

87 j 34b

^{59b 31c 49c}

53b

^{40 b 48 c 45c}

53c 2>c

59c 54a

32c 38b

57b 2 9 b

39b 52c

34c 4 2 c 36 c 6 0 c 27b 3 3 c 35b 5 0 c 27 c 2 6 c 63b 65c 54b 6 5 c 61 c

37 c

54 c 40 c 39 c 6 2c 4 5 c 57c 35c 64b 6 3 c 29c 64c 3 0c 35c ^{2 2}a ^{2 0}a 24a 1 8 a i $ a 87k 2 3 a ^{2 2}b 19b ^{2 1}a ^{2 0}b 1 8b 24b 1 5a 1 8 c 19c 1 4 a 17 a 24 c 1 5 b 15a 23b 1 8 b 14b 2 0c 671 15c 17b ^{1 4}c 57m 2 1 b 23 c ^{2 2}c 2 1c 1 5 c 17c 67n Spaces denote blocks that were not filled by test varieties

Figure 4.3 1979 Trial - Arrangement of varieties on plot

95a 37a 1 1 2a 99a 124a 11 5a ^{8 4}a 97a l2 7 a ^{8 8}a 13 1 a 59 a 1 2 5 a 8 7a 75a 12 9a l0 7 a ^{1 0 1}a ^{9 1}a 57a ^{7 1}a 8 3a 72a 1 2 1a 1 0 5 al 1 2 0a 1 1 1a 1 2 5 a 85a 39a 117a 9 3 a 1 2 3a ^{8 8}a 58 a 89a 1 1 4a 5 2 a 1 0 4a 74a l a 98a 67b ^{1 0 2}a 1 1 9a 1 5 a 1 2 8 a 62a 78a 75a 1 1 3a 1 1 8 a 8 2 a ^{1 1 6}a 29a 1 0 8 a 77a 79a ^{1 0 0}a 8l a 73a 96a 1 0 9a 1 3 0 a | l l 0 a 8 0 a 94a ^{1 2 2}a 1 8 a 1 0 8 a 92a 45a ^{9 0}a 87 c 58a 1 0 3 a 70 a 7 1 b 1 2 4b 84 b 37b 95b 89b 1 1 2b 87b 8 8b 1 0 4 b 85b l b 1 0 5 b ^{1 0 2}b ^{1 1 0}b 67 d ^{1 0 0}b 75b 1 3 0 b 82b 97b ^{9 2}b 74b 8 2 b 1 2 2b

29b 1 3 1b ^{7 0}b 1 1 1b 125b 7 3 b 67e 113b ^{7 8}b 9 3 b 1 0 1b 4 5 b 58 b 77b &5b 69b ^{1 2 0}b 55b 5 2 b 18b 8 0 b 103b 107b 99b 8 1b 15b 8 3 b 1 2 3 b 91b 90b I 2 l b 117b 105b 108b 125b 127b ^{9 8}b 128 b 72b 57 f 1 1 4 b 79b 1 1 8 b 1 1 5 b 75b 39b l l 9 b 94b 115b ^{1 0 9}b 1 2 9b 951b 1 2 1c ^{8 8}c 15c 1 1 3 c 29c 74c ^{1 3 1}c 93c 95c ^{8 4}c 75c 57 Ç 115c 1 09 c ^{8 6}c 95c ^{7 2}c 99c 97 c 52c 80 c 7 1c 18c llO c 1 0 5 c 75c 94 c 1 2 7 c 3 2 c ^{1 2 2}c 58 c 37 c ^{8 2}c 1 0 4 c 57h 5 8 c 1 1 1c 9 2c lOOc 115c 1 2 5 c ^{9 0}c I c 78 c 69 c 79c 7 7 c 1 24c ^{9 8}c ^{1 1 2}c 70 c 1 0 2c 1 2 8 c 57 j ^{1 0 5}c 87c ^{1 0 1}c 39 c ^{1 2 0}c 8 5c 81c 1 1 7c 89c 1 2 3 c 1 1 8c 1 1 9 c 1 3 0 c lo 7 c i i 4 c 1 0 3 c I 2 9 T 73c 45c ^{9 1}c 83c l 0 8 c l 2 5 c

Spaces denote blocks that were not filled by test varieties

ii disease severity — where the area of plant tissue affected is expressed as a percentage of the total area (James, 1974),

There is no straightforward assessment that is appropriate for all plant diseases. Large (1966) recommended a general strategy of

investigation which concluded in the production of a standard diagram for each disease. Although a photograph and a standard diagram were used in PFc.ren.b years of the experiment, the five point scale of 1978 is strictly comparable to the one used in 1979.

The photograph (Figure 4.4) shows leaves selected to illustrate the levels of infection used to assess the disease severity during the

summer of 1970, Each leaf was matched against the photograph to determine the level of infection.

Score Description of infection 1 — No infection

2 — Uredia minute to small, distinct and scattered 3 — Uredia small to medium

4 - Uredia medium, vigorous but not compound 5 — Uredia vigorous and compound

This is an arbitary scale and whilst many such indices and rating systems have been used in the past, percentage scales are generally preferred. The latter have the advantage that they standardize assessments and allow comparable assessments to be made by different workers. For this reason the scale used for scoring the rust in 1979 was a modified version of the twelve point logarithmic scale described by Horsfall and Barratt (1945). According to the Ueber - Fechner Law visual grading progresses logarithmically, Horsfall and Barratt

accounted for this limitation in perception by the human eye in their grading system. They also noted that when assessing percentage

disease, the eye assesses the diseased tissue up to 50% and the healthy area above 50%.

The grades in any standard diagram must be clearly distinguishable by eye, therefore the twelve points used by Horsfall and Barratt have been reduced to five in Figure 4.5. No grades are given for greater than 50% disease because this value represents maximum cover. Standard diagrams usually show only the area covered by lesions or pustules, and therefore the grading does not assess the total diseased tissue.

Score % Cover

1 0

2 1 - 8

3 8 - 2 5

4 2 5 - 5 0

5 >50

In both the 1970 and 1979 plot experiments, the susceptibility of each variety was assessed by scoring five leaves selected at random on each of the test plants* The results were recorded on summary tables;

an example for two varieties is shown in Table 4.1. In cases where a plant died the missing data can be estimated by the formula

eT + bB - S (a-l)(b-l) where a ■ number of treatments

b ■ number of blocks

T • sum of items with same treatment as missing item B ■ sum of items in same block as missing item

5 ■ sum of all observed items

where only one item is missing, a reduction in the number of degrees of freedom gives an approximate analysis of variance (Snedecor, 1955, Pg.310).

An index figure for the susceptibility of each variety may be derived from these grades. There are two types of mean value commonly used. The arithmetic mean, obtained by multiplying the number of leaves in each category by the mean percentage of that category, adding the products and dividing by the number of leaves; this mean value,

however, is distorted by extreme individual scores consequently a better index figure is the geometric mean obtained by adding all the individual scores and dividing by the total number of leaves. This second mean value is the one used in the subsequent analysis of results.

Figure 4,4 Photograph showing leaves selected to illustrate the five levels of infection in 1978.

4

Figure 4,5

Percentage cover

Standard diagram for scoring rust infection on antirrhinum leaves in M79.

25

50

' O'

00

⁽

O O O

4->

4, Quantitative Techniques a ) Design of the experiment

i The variation and interaction of variables within each variety.

The statistical analysis of both the completely randomized and randomized block designs is straightforward with an analysis of variance. The BMD02V computer programme (1977) was used to compute an analysis of variance for factorial designs. The output of this programme includes an analysis of variance table and a grand mean.

Tables 4.2 a end b and Tables 4.3 a and b show samples of the output for variety number 90 - Rocket Red'and number 91 -*Rocket Orange in the first scoring at both plots in 1979. The values for F were calculated for a three factor analysis of variance with variables 1 and 2 (plants and leaves) fixed and variable 3(blocks) random (Zar, 1974). The significance of the calculated values of F was determined by comparing them with the tabulated critical values given by Rohlf and Sokal (1969). Three significance levels were used as shown;

0*05 * significant

0"01 ** highly significant 0*001 *** very highly significant

ii Uniform distribution of rust over each plot.

Table 4,4 shows a sample of the output of the BflDOZV three factor analysis of variance for the fourteen blocks of the control variety "Malmaison” at the first scoring at Wisley in 1978, In addition, the mean score for rust severity was determined for each block of nine plants. The values for the three blocks of each variety were compared and the highest scores, intermediate scores and lowest scores plotted on a map. The coefficient of variation was used to assess the variation of blocks within one variety. This value, which is generally expressed as a percentage, is obtained by dividing the sample standard deviation by the sample mean.

iii Comparison of the epidemic of both localities.

The starting date and the severity of rust infection at both locations were compared for both years of the experiment.

Table 4,2 Sample of computer output of Analysis of Variance within each variety in the first scoring at Royal Holloway College in 1979

a) Variety Number 90 - Rocket Red

B^D02V - ANALYSIS OF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1055 HEALTH SCIENCES CONFUTING FACILITY. UCLA

BND02V - ANALYSIS OF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1955 HEALTH SCIENCES CONFUTING FACILITY. UCLA

Table 4,3 Sample of Computer output of Analysis of Variance within each variety in the first scoring at Wisley in 1979

a) Variety Number 90 - Rocket Red

BNDD2V - ANALYSIS OF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1955 HEALTH SCIENCES COMPUTING FACILITY. UCLA ,

BND02V - ANALYSIS OF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1955 HEALTH SCIENCES COMPUTING FACILITY. UCLA

Table 4.4 Sample of Computer output of Analysis of Variance for the fourteen blocks of "Malmaison" in the first scoring at Wisley in 1979

BMO02V - ANALYSIS QF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1955 HEALTH SCIENCES COMPUTING FACILITY . UCLA

PROBLEM NO. UI

NUMBER OF VARIABLES 3

NUMBER OF REPLICATES 1

VARIABLE NO . OF LEVELS

1 9

2 5

3 14

(14F5.2)

GRAND MEAN 3.93551

SOURCE OF DEGREES OF SUMS OF MEAN

VARIATION FREEDOM SQUARES SQUARES

1 8 .77460 .09583

2 4 .30159 .07540

3 13 158.21587 12.17045

12 32 15.52698 .46522

13 104 54.06984 .51990

23 52 15.72063 .32155

RESIDUAL 415 205.85079 .49463

TOTAL 629 451.46032

b) Assessment of seasonal variation on rust susceptibility.

Ideally agricultural and horticultural experiments should be

performed at a number of places and over a number of years. The effect of the environment on the epidemic may vary considerably from year to year and the purpose of repetition is to determine whether it is possible to make recommendations which are widely applicable.

Yates and Cochran (1938) devised a method to compare trials in

different years using the mean square for experimental error. If all the elements of error are essentially the same the errors may be pooled end the mean of the error mean squares is called the true error variance.

The error mean squares when divided by the true error variance and multiplied by the number of degrees of freedom will be distributed as

Since the antirrhinum trial was only repeated once at two locations, the sample was too small for the method described by Yates end Cochran.

Therefore, the ten varieties included in both years of the experiment were

compared with Kendall's Coefficient of Concordance, ' W , (Siegel, 1956) where 'W measures the extent of association among k sets of rankings of N entities

W ■ S

(n3 - N)

in which S ■ (Rj - Rj)

k ■ number of sets of rankings N ■ number of items ranked

(N — N) ■ maximum possible number of squared deviations

The Null Hypothesis (Ho) is that the k sets of rankings are independent. When N is greater than 7 the expression given below is approximately distributed as X ^

“Y ^ 1--- • k (N — l) W with (N — 1 ) degrees of freedom k N (N + 1)

if the calculated value for is greater than the tabulated value, the null hypothesis is rejected.

c) Comparison of Mean Score for each Variety

The BMD02V programme may also be used to compute a four factor analysis of variance for all the varieties in one plot. An example of this output is given in Table 4,5. The values for F were calculated with variables 1, 2 and 3 (variety, plants and leaves) fixed and variable 4 (blocks) random (Zar, 1974).

The four factogènalysis of variance table can only show whether there are significant differences between some of the varieties. There are several methods for locating these differences and in this case the method devised by 3.W. Tukey and described with some modifications by Snedecor (1956) has been used. This test uses the experiment as a whole for assessing the risk of error and this value may be calculated from the Mean Square of the Residual Error

. _ ? Mean Square e.g. Variance Sx *---

---. ---.---. „ / Mean Square Standard Deviation Sx « / ---- ---Standard Error • Sx

The test may best be explained by an example. The mean scores for each variety was ranked as shown in the left hand column of Table 4.6.

Table 4,5 Sample of computer output of a four factor Analysis of Variance for all the varieties in the second scoring at Royal Holloway College in 1979,

BMD02V - ANALYSIS OF VARIANCE FOR FACTORIAL DESIGN - VERSION OF JULY 22. 1955 HEALTH SCIENCES COMPUTING FACILITY , UCLA

PROBLEM NO. 01

NUMBER OF VARIABLES 4

NUMBER OF REPLICATES 1

VARIABLE NO. OF LEVELS

1 74

2 0

3 5

4 3

(15F5.2)

GRAND MEAN 1.90931

SOURCE OF DEGREES OF SUMS OF MEAN

VARIATION FREEDOM SQUARES SQUARES

1 73 2035.64905 27.89930

2 8 14.08468 1.75059

3 4 44.65505 11.15401

4 2 250.02943 125.01471

12 684 415.10050 .71079

13 292 75.97357 .25351

14 146 1295.57798 8.88057

23 32 7.67908 .23997

24 16 11.32012 .70751

34 8 8.44605 1.05575

123 2336 462.69129 .19807

124 1168 855.07247 .73294

134 584 134.50210 .23031

234 64 13.51882 .21279

RESIDUAL 4672 917.43303 .19537

TOTAL 9989 6545.83423

Each mean is then subtracted from those above to form the half matrix in Table 4.6. The test is made by computing a difference D which is significant at the SJb level and then comparing this value with the differences between the means

D ■ Sx x Q

Q is taken from the Studentized Range (Rohlf and Sokal, 1969) In this case D ■ 0*0482 and if the difference between the means is greater than the value for D the two varieties are significantly different. In Table 4.6 the insignificant differences are underlined,

Table 4.6 Matrix showing significant differences between the first ten varieties in the second scoring at Royal Holloway College in 1979 as distinguished by Tukey's Test

Var•Code 62 131 18 126 122 74 71 128 1

No.* X x-1'09 x-1.12 x-1'21 x-1'21 x-1'21 x-1'33 x-1'34 x-1'37 x-1'37 103 1*38 0*29 0*26 0*17 0*17 0*17 D'05 0'04 G'Ol G'Gl

1 1*37 0-28 0*25 0-16 0*16 0'16 0'G4 0'G3 G'GG 128 1*37 0*28 0*25 0*16 0*16 G'16 G'04 0'G3

71 1*34 0*25 0*22 0*13 0*13 0'13 G'Gl 74 1*33 0*24 0-21 0*12 0"12 0'12

122 1*21 0*12 0*09 0*00 O'GO

126 1*21 0-12 0*09 0*00 D ■ 0'G482

insignificant differences

18 1*21 0*12 0*09 are underlined

131 1*12 0-03

62 1*09

* see Appendix 4.1 for the name and source of each variety or Appendix 4.14 for a fold out list of the variety names.

d) Classification to group varieties with similar resistance.

The insignificant differences shown in the Matrix of Tukey's Test may be used to classify the varieties with respect to their susceptibility

to the rust fungus. The classification produces discrete groups (A, B, C etc.) such that similarities within groups are greater than those between groups.

Figure 4,6 shows an example of the classification on the first ten varieties in the second scoring at Royal Holloway College in 1979.

Figure 4,6 shows an example of the classification on the first ten varieties in the second scoring at Royal Holloway College in 1979.

In document Estudio histórico-constructivo de la (página 77-90)